Closed vlendec closed 6 months ago
Hi Volker, in what case does this happens?
I wonder if we should rather return an error if the oldset is empty.
I'm playing with the gssproxy protocol, and got mechglue to crash. This is not with the upstream gssproxy daemon that this project provides, but I thought it might be worthwhile to not crash unsuspecting clients no matter what happens on the daemon side. See also the SIGPIPE patch of a while ago. With the real gssproxy this should never happen.
I see, would you mind changing the patch to return an error instead of success + empty set ? I think empty set will fix the segfault of course, but will also mask that incorrect (missing) data is being returned.
Sure, but I guess this should happen when parsing the bytes coming from the server. Let's close this PR, I'll take a fresh look.
Protect mechglue against 0-length OID_set coming from the gssproxy